ARTICLE

Received 10 Dec 2013 | Accepted 26 May 2014 | Published 8 Jul 2014 DOI: 10.1038/ncomms5272 OPEN Cdon acts as a Hedgehog decoy receptor during proximal-distal patterning of the

Marcos Julia´n Cardozo1,2, Luisa Sa´nchez-Arrones1,2,A´frica Sandonis1,2, Cristina Sa´nchez-Camacho2,w, Gaia Gestri3, Stephen W. Wilson3, Isabel Guerrero1 & Paola Bovolenta1,2

Patterning of the vertebrate optic vesicle into proximal/optic stalk and distal/neural retina involves midline-derived Hedgehog (Hh) signalling, which promotes stalk specification. In the absence of Hh signalling, the stalks are not specified, causing cyclopia. Recent studies showed that the Cdon forms a heteromeric complex with the Hh receptor Patched 1 (Ptc1). This receptor complex binds Hh and enhances signalling activation, indicating that Cdon positively regulates the pathway. Here we show that in the developing zebrafish and chick optic vesicle, in which and ptc1 are expressed with a complementary pattern, Cdon acts as a negative Hh signalling regulator. Cdon predominantly localizes to the basolateral side of neuroepithelial cells, promotes the enlargement of the neuroepithelial basal end-foot and traps Hh , thereby limiting its dispersion. This Ptc-independent function protects the retinal primordium from Hh activity, defines the stalk/retina boundary and thus the correct proximo-distal patterning of the .

1 Centro de Biologı´a Molecular Severo Ochoa, Consejo Superior de Investigaciones Cientı´ficas—Universidad Auto´noma de Madrid, 28049 Madrid Spain. 2 Ciber de Enfermedades Raras (CIBERER), ISCIII, c/ Nicolas Cabrera 1, 28049 Madrid Spain. 3 Department of Cell and Developmental Biology, University College London, Gower Street, London WC1E 6BT, UK. w Present address: Universidad Europea, C/ Tajo, s/n Villaviciosa de Odo´n, 28670 Madrid, Spain. Correspondence and requests for materials should be addressed to P.B. (email: [email protected]).

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 1 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272

he Hedgehog (Hh) signalling pathway is a highly expression, whereas ptc1 matches Pax2 distribution26,34,35 conserved regulator of patterning and homeostatic events (Fig. 1f–i; Supplementary Fig. 1). Here we have addressed Tin various developing and adult organs and thus its whether and how cdon contributes to optic vesicle patterning, activation is under tight control1,2. Hh ligands are released using zebrafish and chick embryos as model systems. Our results through the cooperative activity of the transmembrane protein indicate that in the eye, Cdon acts as a negative regulator of Hh Dispatched and heparan sulphate proteoglycans, which are also signalling. Cdon as well as Boc expand the surface of the basal crucial to modulate long-range distribution3–5. There is also end-foot of neuroepithelial cells, bind Hh and prevent its evidence that cytonemes, long filopodial-like extensions formed dispersion, thus limiting long-range signalling, as proposed for by the producing cells, promote delivery of Hh-containing Ihog and Boi19,36–38. This function protects the retinal vesicles to similar extensions in target cells6–8. In the targets, primordium from Hh activity, thus enabling a correct P-D binding of Hh to a receptor complex containing the seven-pass patterning of the eye. membrane protein Patched (Ptc), releases Ptc-mediated inhibition of the signal transducer protein Smoothened (Smo), Results which initiates a cascade of intracellular events that lead to the 9,10 cdon and ptc localize to complementary optic vesicle domains. transcription of Hh target . Pax2 function is critical to establish optic stalk identity in dif- Cdon (cell adhesion molecule-related, downregulated by 32,33,39 11,12 13 ferent vertebrate species and to restrain Pax6 expression, so oncogenes) , the related Boc (Brother of Cdo) or their that the two genes have a complementary expression restricted to Drosophila counterparts Ihog (Interference hedgehog) and Boi the proximal/stalk and distal/retinal optic vesicle, respectively29,32 (Brother of Ihog), are additional components of the Hh receptor (Supplementary Fig. 1). Pax2 expression is regulated by Hh, complex14. Cdon and Boc are cell surface glycoproteins that which is strongly expressed in the ventral midline30,32,33. To get a belong to a subgroup of the immunoglobulin (Ig) super-family of better understanding of the role of Hh signalling in the P-D cell adhesion molecules14. Their ectodomains contain five and patterning of the eye, we analysed the distribution of Hh in four immunoglobulin-like domains, respectively, followed by zebrafish embryos. Hh was strongly localized to cells of the three fibronectin III (FNIII) repeats (1–3), which are involved in 15 ventral midline of the forebrain and in optic stalk cells (Fig. 1a–e). Ptc and Hh binding. FnIII(1-2) domains interact with Ptc , At early stages of optic vesicle specification, Hh strongly whereas the evolutionary conserved FnIII(3) domain binds Hh 16–18 accumulated at the most ventrally positioned Pax2-positive cells with high affinity . (Fig. 1c) but its distribution extended dorsolaterally as develop- The requirement of a heteromeric interaction between Ptc and ment progressed (Fig. 1d,e). In agreement with this distribution, Ihog/Boi for both high-affinity Hh binding and presentation of 19 the Hh receptor ptc1 was expressed in the ventral forebrain Ptc at the cell surface was initially demonstrated in Drosophila 34,40 15 including the prospective optic stalk (Fig. 1f,g). In contrast, and then shown to be conserved in vertebrates .Inmice, cdon, but not the related boc28 (Supplementary Fig. 1w–z), was Cdon and Boc and the unrelated Hh binding protein Gas1 strongly localized to the neural retina region (Fig. 1h,i). This (refs 20,21) act as positive and redundant regulators of Hh distribution was conserved in chick and mouse embryos signalling in different contexts22–24.Forexample,Cdon-deficient (Supplementary Fig. 1k–o). In the latter, Cdon protein appeared mice are characterized by mild holoprosencephalia (HPE) with to accumulate at the basolateral membrane of the retinal microphthalmia and moderate facial and brain defects of strain 11,25–27 epithelium starting from E10, resembling the subcellular specific penetrance . These traits are aggravated by the 4,36 À / À À / À localization described for the Drosophila homologue Ihog additional inactivation of Boc,sothatCdon ;Boc embryos (Supplementary Fig. 1t–v). present a significant loss of the Hh pathway activity24,whichis further worsened by Gas1 inactivation. Indeed Cdon À / À ; Boc À / À ;Gas1 À / À compound mutant embryos show a cdon knockdown alters P-D patterning of the eye primordium. phenotype similar to that caused by the loss of The complementary pattern of ptc1 and cdon expression was (Shh) function: strong facial malformations and failure of partition consistent with the idea that, in the eye, the two could act of the cerebral hemispheres and eye primordia22.Inzebrafish, independently without forming a heteromeric Hh receptor15, analysis of the uml() mutant supports the idea that Boc acts as a raising the question of whether and how cdon expression in the positive regulator of Hh signalling in the ventral but retina contributes to . Answering these might have an opposite function in the lower jaw28. questions, however, requires local or partial interference with In vertebrates, the morphogenesis of the eye primordium, also Cdon function because shh colocalizes with cdon at the axial known as optic vesicle, occurs concomitant to its patterning midline of gastrulating embryos (Fig. 1f,g). To avoid the complete along the proximal-distal (P-D) axis in medial (prospective optic abrogation of this early function that may be responsible for the stalk) and lateral (prospective retinal) domains. The establish- mild holoprosencephalic phenotype of Cdon null mouse ment of this early P-D patterning is defined by the expression of embryos11, we reduced cdon expression levels in zebrafish two pair- and homeobox- containing transcription factors, Pax2 embryos using specific (MOs). and Pax6, the expression of which is, respectively, restricted to the Doses of 160 mM of a MO that efficiently interfered with the optic stalk and the retinal primordium29–31. Functional studies cdon translation start site (CdonATG MO; Supplementary Fig. 2a–e) have demonstrated that Hh secreted from the midline of the generated morphants, which were characterized by abnormal ventral forebrain is required to promote Pax2 expression and folding of the retinal ventronasal quadrant and the presence of an thus to establish optic stalk identity30,32,33. A cross-repressing optic fissure coloboma (lack of fissure fusion; Supplementary regulatory loop between Pax2 and Pax6 then refines the pattern, Fig. 2f–i; higher doses led to early lethality), often a consequence of forming a precise boundary between the retina and the optic abnormal optic stalk development31,41,42. The trunk in morphants, stalk29. In the absence of Hh signalling, the stalk does not form however, appeared normal with no apparent ventral midline and Pax6-positive vesicles do not separate, forming a single defects suggesting that cdon knockdown was compatible with axial cyclopic eye33. midline formation. These anterior defects were not found in Notably, in the optic vesicles ptc and cdon are distributed embryos injected with a control MO (MOct) nor were rescued with a complementary and evolutionary conserved pattern: by the co-injection of a MO against p53 (n ¼ 52), which is Cdon localizes to the distal region overlapping with Pax6 activated by MO injection and can contribute to off-target effects43.

2 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE

Shh/Pax2 Shh/Pax2 ptc1 cdon optic vesicle stage 20 ss a p ov ov

os p os

a Shh/Pax2

os

Shh/Pax2 stage

cdon shh 75%–90% Epiboly d

os v Shh/Pax2

Figure 1 | Expression of cdon and other Hh signalling components during P-D patterning of the optic vesicles. (a–c) Ventral views of anterior zebrafish forebrain at optic vesicle (a is ventral to b) and optic cup (d, e) stages immunostained in toto with against Pax2 and Shh. (f–i) Dorsal (f,h) and lateral (g,i) views of zebrafish embryos hybridized in toto with probes specific for ptc1 and cdon at the 20 ’ stage (20 ss). (j–k) Coronal sections of embryos (75–90% epiboly) hybridized for cdon and shh. The white arrows in b,c and e point to Shh immunolabelling in Pax2-positive cells. Dashed lines in a–d,j and k indicate the embryonic midline. Continuous white and black lines in f–i outline the lens and the embryonic border, respectively. a, anterior; d, dorsal; os, optic stalk; ov, optic vesicle; p, posterior; v, ventral. Scale bar, 25 mm.

Injection of a different cdon MO complementary to the 30 splice Altogether these data support the idea that cdon is specifically site of exon 8 (Cdonspl8 MO) phenocopied cdonATG morphants required for restraining the size of the optic stalk and ventral (Supplementary Fig. 3a–b). retina. To investigate more precisely the developmental nature of these defects, we analysed the expression pattern of well- established markers of the optic stalk, pax2.1 and fgf8a32,44.In cdon counteracts Hh signalling during P-D eye patterning.In control embryos, pax2.1 was expressed in the optic stalk, mammals, Cdon acts as a positive modulator of Hh signalling in midbrain–hindbrain boundary (MHB) and along the hindbrain the midline and its inactivation in mice causes a microform of (Fig. 2a,c,e). In most cdonATG and cdonspl8 morphants, pax2.1 was HPE11,15 associate with a decrease in the Pax2-positive optic stalk enlarged invading a portion of the ventral optic cup but reduced domain26, a phenotype opposite to that observed in cdon in the hindbrain (124/140 injected embryos; Fig. 2b,d,e; morphants. As a possible interpretation of this discrepancy, we Supplementary Fig. 3c–d). In lateral views of whole-mounted postulated that the mouse optic vesicle phenotype might be a cdonATG morphants, there was an approximate doubling of the consequence of an early Cdon function, which is instead visible area of the optic stalks and ventral retina that expressed preserved in the knocked-down cdonATG morphants, unveiling pax2.1 (0.007 mm2±0.0003 s.e.m; n ¼ 45) as compared with a later and tissue-specific cdon function. control embryos (0.0033 mm2±0.0002 s.e.m; n ¼ 28; T-test, This function was consistent with Cdon acting as a negative Po0.001). A similar expansion of the optic stalk in cdonATG modulator of Hh signalling. If this were the case, inactivation of morphants was observed also with an fgf8a probe specific (93/112 Hh signalling should rescue the cdon loss of function phenotype. injected embryos; Fig. 2f–j). fgf8a distribution was expanded To test this hypothesis, we interfered with Hh signalling in also in MHB, dorsal diencephalon and telencephalon (Fig. 2f–j). cdonATG morphants by incubating the embryos in a medium Because coloboma is often associated with ventral retina containing the Smo antagonist cyclopamine. As expected, this defects, we also analysed the expression of two ventral retina treatment efficiently abolished the expression of pax2.1 in the specific markers, pdzrn3 (ref. 45) and rdh10a (retinol optic stalk of control embryos (Fig. 3a,c). Notably, cyclopamine dehydrogenase 10a)46. Their distribution was expanded in also counteracted the expansion of pax2.1 in the optic stalk of cdonATG morphants in comparison with that of control CdonATG morphants, generating embryos with an overtly wild- embryos (Supplementary Fig. 3e–h) but this increase was not type phenotype (34/35 embryos; Fig. 3a,b,d). associated with an appreciable reduction of the expression of the Together these experiments point to the idea that Cdon, dorsal retinal marker tbx5.1 47 (Supplementary Fig. 3i,j). when not properly matched by Ptc presence, may act as a

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 3 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272

Control cdonATG MO Control cdonATG MO ab fg

] ]

c dhi fgf8a pax2.1 ] ]

e mes j di MHB di mes tel tel MHB rh os Eye os Eye rh

Figure 2 | The optic stalk is expanded in cdon morphants. (a–j) In situ hybridization analysis for two optic stalk markers, pax2.1 (a–d) and fgf8a (f–i) at 28 hpf and 24 hpf, respectively. Embryos are shown in lateral (a,b,f,g) and frontal (c,d,h,i) views. Expression patterns of both genes are schematically represented in e and j.IncdonATG morphants, pax2.1 expression is expanded dorsally in the optic stalk (b,d brackets) when compared with controls (a,c brackets). Fgf8a expression is expanded caudally and laterally in the optic stalk (g,i arrowhead and dotted lines) as well as in the telencephalon (g,i) when compared with control embryos (f,h arrowhead and dotted lines). di, diencephalon; mes, mesencephalon; MHB, midbrain–hindbrain boundary; os, optic stalk; rh, rhombencephalon; tel, telencephalon. Scale bars, 100 mm.

Control cdon ATG MO observed in cdon morphants as compared with control embryos (Supplementary Fig. 3k–o). MHB

Cdon function in the eye depends on the Hh binding domain. Studies in vertebrates have suggested that Cdon interacts with Ptc1 through FnIII(1-2) domains23, forming a heteromeric DMSO receptor15. Cdon further binds Hh through the remaining FnIII(3) domain16,18. Alignment of the zebrafish, mouse and os human CDON sequence revealed conservation of the residues pax2.1 pax2.1 defined as relevant for Ptc and Hh binding15,16 (Supplementary Figs 4–7). Furthermore, analysis of the zebrafish cdon genomic (NC_007129.5) and protein (Q1L8D0) sequences indicated that it was possible to design specific splicing MOs, which would cause skipping of exon 11 or 14 without altering the protein-reading frame (Fig. 4a,b). These MOs, designed as cdonspl8, at the exon- intron or intron-exon boundaries and named cdonspl14, cdonspl11aand cdonspl11d, injected alone (cdonspl14)orin Cyclopamine combination (cdonspl11aand cdonspl11d) efficiently generated mutated Cdon versions, considerably reducing the wild-type pax2.1 pax2.1 cdon form (Fig. 4). The prevalent mutated forms were, as expected, a frame-shift mutation (cdonspl8) or forms that lacked Figure 3 | Cdon acts as a negative modulator of Hh signalling. the functionally relevant sequences of FnIII(3) (cdonspl14) and (a–d) Lateral views of control or cdonATG morphants treated with DMSO FnIII(2) (cdonspl11aand cdonspl11d) domains, as determined by (vehicle) or cyclopamine from 90% epiboly and analysed with ISH for PCR with reverse transcription (RT–PCR) analysis of the pax2.1 expression at 28 hpf. Blocking Hh signalling abolished pax2.1 respective morphants (Fig. 4c,f,i), followed by sequencing of the expression in the optic stalk in wt embryos (c). In cdonATG morphants, corresponding bands (Supplementary Fig. 8). Furthermore, cyclopamine treatment counteracts the expansion of pax2.1 overexpression transfection of similar constructs into HEK cells demonstrated observed in cdonATGMO injected embryos (a,b,d). The lens and the that the deleted versions of the protein were efficiently inserted body are outlined with black and white dashed lines in (a–d). Arrows into the plasma membrane (Supplementary Fig. 9). We thus indicate the extent of the pax2.1 expression in the optic stalk. DMSO, expected that cdonspl11a/spl11d and cdonspl14 MO injections would dimethyl sulphoxide; MHB, midbrain–hindbrain boundary; oc, optic cup; os, generate embryos carrying Cdon versions with a reduced optic stalk. Scale bar, 100 mm. capability of interacting with either Ptc1 or Hh. As observed in cdonATGMO and cdonspl8 injected embryos (Figs 2a–e and 4c–e,l), cdonsp14 morphants (58/86) were negative modulator of Hh signalling in the optic vesicle. This characterized by a significant expansion of the pax2.1-positive idea was also supported by the expansion of the expression optic stalk domain, phenocopying the cdonATG phenotype domain of nkx2.2, a ventral forebrain Hh target gene28, (Fig. 4f–h,l). In contrast, cdonsp11a/ spl11d morphants (110/110)

4 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE resembled wild-type embryos with no changes in the extent of the was localized in the ventral neuroepithelium with a symmetric pax2.1 domain despite the evident splicing effect generated by the distribution that covered about 125 mm from the midline. This MOs co-injection (Fig. 4i–k,l). symmetric distribution was no longer observed when Cdon was These results support that, during optic vesicle development, expressed in the adjacent optic stalk neuroepithelium (25/25 of Cdon function depends on Hh binding but likely with a Ptc- positive embryos). In this case, the ectopic Cdon-positive cells independent mechanism. If this were the case, overexpression of closest to the Hh source presented an accumulation of Hh Cdon or CdonDFnIII(2) should rescue the phenotype of cdon immunoreactivity, particularly evident at the basolateral side of morphants, whereas CdonDFnIII(3) should have no effect. To test the cells. This strong signal was mostly limited to the Cdon- this hypothesis we took advantage of the tg(rx3::Gal4) line48. This positive cells proximal to the Shh domain and less evident in line enables UAS-mediated overexpression of cdon only in the Cdon-positive cells that occupied a position more distal to the Rx3-positive domain (Supplementary Fig. 10a–c), thus offering Shh source (Fig. 6e–h; Supplementary Fig. 12d–f,m,n). Notably, the possibility of further testing if the zebrafish morphant the effect of Cdon on Hh protein distribution was independent of phenotype is linked to retinal-specific cdon function. Injection of Cdon interaction with Ptc1 because ectopic expression of UAS::GFP DNA followed by a slightly delayed injection of CdonDFnIII(1-2) caused a similar predominant accumulation cdonATG MO efficiently drove GFP expression in the developing of Hh at the most proximal Cdon-positive cells (Fig. 6i–l; optic vesicle (Supplementary Fig. 10c) but did not rescue the Supplementary Fig. 12g–i,m,n). On the contrary, forced expansion of the pax2.1 expression domain caused by cdon expression of CdonDFnIII(3) had little effect on the symmetric knockdown (Supplementary Fig. 10f). UAS::Cdon and Hh distribution (Fig. 6m–p; Supplementary Fig. 12j–n), although UAS::CdonDFNIII(2), but not UAS::CdonDFNIII(3) DNA, proximal cells expressing this Cdon mutated version appeared to instead significantly counteracted the expansion of the pax2.1- bind more ligand than control cells, opening the possibility that positive optic stalk domain characteristic of the morphants Cdon residues other than those present in the FNIII(3) domain (Supplementary Fig. 10d–f). partially contribute to Shh binding. Alternatively, this Cdon Together these data support that Cdon expression in the retinal deleted form may enhance the Hh binding efficiency of other field is required to establish the optic stalk boundary with an Hh receptors expressed in the neuroepithelium. dependent but Ptc-binding independent mechanism. Notably, Cdon-positive neuroepithelial cells presented a significantly enlarged basal end-foot as compared with Cherry electroporated control cells (Fig. 7a–f,s; Supplementary Retinal Cdon expression preserves the retina/stalk boundary.In Fig. 13a–f). Furthermore, in control cells the end-foot was rather Drosophila both Ihog and Boi can sequester the amount of Hh simple, whereas in Cdon-positive cells the end-foot was decorated available thereby limiting long-range signalling19,36–38. We thus by several filopodial-like extensions that appeared to be a main hypothesized that Cdon expressed in the retinal primordium site of Hh accumulation (Fig. 7d–f; Supplementary Fig. 13d–f). could bind Hh preventing its dispersion into the future retinal Consistent with the idea that Cdon acts as a ‘sink’ for Hh protein, domain. In this case, loss of Cdon function should expose retinal cells immediately surrounding a Cdon-positive cell were devoid cells to Hh, thereby expanding the stalk domain. of Hh immunoreactivity (Fig. 7e,f; Supplementary Fig. 13e,f). A To test this possibility, we interfered with Cdon expression in a similar significant enlargement of the basal end-foot associated spatiotemporal controlled manner, taking advantage of the with the presence of some cytoplasmic extensions was observed evolutionary conserved distribution of Cdon in the chick optic also in cells that ectopically expressed CdonDFnIII(1-2) vesicles (Supplementary Fig. 1o,s). A cCdonATG-tagged-MO was (Fig. 7g–i,s; Supplementary Fig. 13g–i), but not in those carrying forced by focal in ovo electroporation into the eye anlage of HH8 the CdonDFnIII(3) protein (Fig. 7j–l,s; Supplementary Fig. 13j–l). embryos, when Cdon expression in vesicle appears Together these results indicate that Cdon influence of basal end- (Supplementary Fig. 11). Targeted embryos showed an expansion foot morphology is independent from Cdon interaction with Ptc of pax2 expression, which was not observed in embryos but is likely linked to Hh interaction. electroporated with an equally tagged cMOct or in the Similar results were observed when the related Boc-EGFP was contralateral eye of unilaterally cCdonATG electroporated ectopically expressed in the posterior neural tube supporting the embryos (Fig. 5a–d,f). This phenotype resembled that of gain of idea that the two proteins can act in a similar manner23,24. Boc- midline-derived Hh signalling32,49. Expansion of pax2 expression positive neuroepithelial cells accumulated Shh (not shown) and was associated with a clear reduction of Pax6 distribution in the presented an enlarged basal end-foot decorated by several optic cup (Fig. 5e,g), well in agreement with the notion that Pax2 filopodial-like extensions (Fig. 7m–r). This end-foot was larger and Pax6 regulate each other’s activity29. (6.2 mm ±0.7 s.e.m; n ¼ 7; P 0.001) than that of GFP-positive Together these data indicate that Cdon expression in the retinal r neuroepithelial cells (2.6 mm±0.3 s.e.m.; n ¼ 6). primordium controls the balance between Pax2/Pax6 thereby All in all, these results support the idea that Cdon and Boc establishing the boundary between the optic stalk and the optic localize predominantly at the basal surface of the neuroepithelial cup. Considering that Pax2 is an Hh target30,32,33, the position of cells enriched in filopodial-like extensions. These protrusions may this border depends upon Hh protein distribution, which, in turn, serve to enhance Hh binding in Cdon/Boc expressing cells, thus may be influenced by Cdon acting as a barrier. modifying Hh distribution and thereby its activity.

Cdon controls Hh dispersion in the neuroepithelium. We next investigated if Cdon expression modifies the distribution of the Discussion endogenous Hh protein from the ventral hypothalamus. To this Our study provides evidence that Cdon-mediated interference end, we used in ovo electroporation and expressed mouse Cdon with Hh ligand dispersion is a mechanism by which Hh signalling and its deleted derivatives17 ectopically into the optic stalk of information can be regulated in vertebrates. We show that Cdon HH8 chick embryos, close to the Hh producing ventral midline. expression in the retinal neuroepithelium serves to prevent the Successfully targeted embryos were then co-immunostained with activation of the Hh target pax2, thereby controlling the antibodies against Cdon and Shh. In embryos electroporated with proximal-distal patterning of the optic vesicle. We propose that a control membrane Cherry-expressing vector (Fig. 6a–d) or in mechanistically this occurs because, in the absence of Ptc co- the non-electroporated half of the embryo (Fig. 6e–h), Hh signal expression, Cdon acts as a sink for Hh proteins that

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 5 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272

RT–PCR Pre-mRNA Exon Exon Exon Exon Non-skipped

Exon Exon Exon Exon mRNA RT-PCR

MOspl Exon Exon Exon Skipped

cdonspl11a zCdon mRNA spl8 spl11d spl14 ATG cdon cdon cdon UGA

Protein FNIII 1 FNIII 2 FNIII 3 Patched binding Hh binding

Control cdonspl8 MO MWM (bp) Control cdonspl8 MO

1,000 750 11 2 500 3

4

250 pax2.1 pax2.1

Control cdonspl14 MO MWM (bp) Control cdonspl14 MO

1,000 750

500 5 5 250 6

pax2.1 pax2.1

cdonspl11a MO Control cdonspl11a MO + cdonspl11d MO MWM (bp) Control cdonspl11d MO 1,000

750

500 77 8

9 250 pax2.1 pax2.1

2.5

*** 2 *** *** 1.5

P=0.431 1 positive optic stalk area

0.5 pax2.1 28 45 12 44 26 65 22 57 0

MO MO MO MO ATG spl8 MO Control Control Controlspl11a Controlspl14 spl1d cdon cdon cdon cdon cdon

6 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE

Figure 4 | Cdon interaction with Ptc is dispensable for its function in the optic vesicle. (a) Schematic diagram of the design of the exon skipping MO used in the study. (b) Schema representing the zebrafish cdon mRNA (zcdon) aligned with the corresponding Cdon protein indicating the domains targeted by cdonspl8, cdonspl11a, cdonspl11d and cdonspl14 MOs. The exons that encode the 50 and 30 UTR regions are depicted in blue; those that, when skipped, generate a translational frame shift are indicated in red; whereas those that, when skipped, maintain the reading frame are indicated in green. (c,f,i) RT–PCR analysis of the exon skipping function of cdonspl8, cdonspl14 and cdonspl11a/cdonspl11d MOs. For detailed information about the resulting bands, noted 1–9, please refer to Supplementary Fig. 8. (d,e,g,h,j,k) ISH analysis of pax2.1 expression pattern in cdonspl8 (d,e), cdonspl14(g,h) and cdonspl11a/cdonspl11d (j,k) MO injected embryos at 26–28 h.p.f. Pax2.1 expression domain is expanded in cdonspl8 (d,e) and cdonspl14 morphants (g,h)in comparison with their respective controls, whereas there was no difference in the pax2.1 expression domain of control and cdonspl11a/cdonspl11d morphants (j,k). (l) Quantification of the optic stalk pax2.1-positive expression domain in embryos injected with the different MOs at 26–28 h.p.f. (***Po 0.001; Student’s t-test). The number of embryos analysed in each case is indicated in each column and are as follows: CdonATG MO injection (control, n ¼ 28; MO, n ¼ 45); Cdonspl8 MO injection (control, n ¼ 12; MO, n ¼ 44); Cdonspl11a and Cdonspl11d MOs injection (control, n ¼ 26; MO, n ¼ 65); Cdonspl14 MO injection (control, n ¼ 22; MO, n ¼ 57). Error bars represent s.e.m. Scale bars, 100 mm. MWM, molecular weight marker.

cCdonATG MO cCdonATG MO (Control side)

ov

α-Fluorescein/pax2

cCdonATG MO (Experimental side)

ov ov

ov

α-Fluorescein/pax2 α-fluorescein/pax2

Control side Experimental side

ov ov

α-Fluorescein/pax2 α-Pax6 α-Fluorescein/pax2 α-Pax6

Figure 5 | Localized interference with Cdon expression in the optic vesicles expands the distal optic stalk domain. (a–f) Lateral (a,b) and ventral (c) views of the control (a) and experimental (b) optic vesicles (ov) of HH14 chick embryos with unilateral focal electroporation of a carboxyfluorescein conjugated cCdonATG MO at HH8. Embryos were hybridized for pax2 (blue signal) and immunostained with anti-fluorescein antibodies (brown signal) to detect MO distribution (a–d,f). Pax6 immunohistochemistry was performed in cryostat sections of electroporated embryos (e,g). Pax2 expression is expanded to the entire optic vesicle of cCdonATG MO-treated embryos (b,c) in comparison with the non-electroporated control eye (a,c). Pax2 expansion is associated with a reduction of Pax6 distribution in the retina (f,g) in contrast to a wild-type condition (d,e). The optic vesicles are outlined with black (a–c) or white (g) dashed lines. Scale bar, 50 mm. predominantly accumulate at the filopodial-enriched basolateral Our notion of Cdon-mediated regulation of Hh signalling is side of Cdon-expressing neuroepithelial cells. A similar sink supported by the analysis of Cdon effect on Hh protein mechanism, mediated by Boi or Ihog, seems to limit Hh long- distribution and by the consequences of graded and spatiotem- range distribution in Drosophila36–38. Therefore we propose that, poral controlled interference with cdon expression in both in the absence of Ptc interaction, Cdon/Boi/Ihog—and perhaps zebrafish and chick-developing optic vesicle. The eye phenotype Boc28—have an evolutionary conserved function as Hh decoy resulting from this interference was compatible with that receptors. This mechanism of limiting Hh activity acts in parallel observed after gain of Hh signalling function in the optic vesicle, to other, more intensively studied mechanisms, for example, which is characterized by the expansion of the pax2-positive optic negative feedback regulation. A case in point is Hh-mediated stalk domain30–32,51, indicating that Pax2 is an Hh target. control of the expression of the Ptc receptor, which, in turn, Consistent with these observations, cdon knockdown caused an inhibits pathway activation50. expansion of pax2.1 eye expression, which was rescued by

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 7 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272

Cherry Cdon CdonΔFnIII(1-2) CdonΔFnIII(3) a e i m os os os os

α-Cherry fp α-Cdon fp α-Cdon fp α-Cdon fp b f j n

α-Shh α-Shh α-Shh α-Shh c g k o

Merge Merge Merge Merge d h l p

Merge Merge Merge Merge

Figure 6 | Cdon overexpression modifies Shh protein distribution in the optic stalk neuroepithelium. (a–p) Confocal analysis of coronal sections at the level of the optic vesicle of HH14 chick embryos electroporated at HH8 with a construct carrying mCherry (a–d), Cdon (e–h) or its deleted derivatives lacking the Ptc (CdonDFnIII(1-2)) (i–l) or Shh (CdonDFnIII(3)) (m–p) binding domains. Sections were immunostained with antibodies against mCherry (a), Cdon (e,i,m) and Shh (b,f,j,n). Note the localization of Shh protein in Cdon-positive cells—particularly in their basal regions (f,h white arrow)—falling outside of the Shh-expression domain (e–h). A similar accumulation is observed in the presence of the CdonDFnIII(1-2) construct (i–l white arrow) but was hardly detectable in mCherry or CdonDFnIII(3) expressing cells falling outside of the Shh domain (a–d and m–p). The midline is indicated with a yellow dotted line. The normal extent of Shh distribution is indicated with a yellow line, whereas extended Shh localization is indicated in red. In p the arrowhead points to a Cdon-positive cell within the Shh domain, whereas the arrow points to the first Cdon-positive cell falling outside the Shh domain. fp, floor plate; os, optic stalk. Scale bar, 50 mm. interfering with Hh signalling activation and by the localized observations may reflect differential sensitivity of target genes overexpression of Cdon within the eye field. Furthermore, ventral to Hh concentrations56. Ptc, gli2, gli2b, gli3 and nkx2.1 likely retina defects similar to those of cdon morphants have been respond to high midline-derived Hh levels, whereas nkx2.2, given reported in other zebrafish lines, in which Hh signalling is its broad expression, might be more sensitive to local Hh overactivated, for example, as in the blowout34, uta152 and modifications caused by interference with Cdon expression. aussicht lines53,inZic2a morphants49 or after Hh overexpression Cdon has a transient but conserved expression in the in zebrafish30,32 and chick embryos51. prechordal plate mesendoderm, and ventral neural Notably, our findings are also in good agreement with studies tube midline, which all produce Hh proteins. Recent studies have in the teleost fish Astyanax mexicanus, a natural example of the shown that Ihog regulates the release of Hh from the producing consequence that Hh signalling expansion has on eye develop- cells with high levels of Ihog blocking Hh release or transport4,36. ment. In their natural environment, the Astyanax mexicanus At the same time, Ihog favours the formation of cytonemes that exists in two forms: a surface-dwelling river morph and a cave- ensure long-range transport of Hh molecules6 and in the living blind morph (cavefish). In the embryonic cavefish, the receiving cells increases Hh signalling19. In a similar manner, ventral expression domains of Hh and tiggy-winkle (twhh), a Cdon may control Hh dispersion shaping the gradient family member that shares Hh functions in the developing eye30, responsible for the specification of ventral fates, including are expanded in the forebrain54,55, Hh target genes, including among them the optic stalk. Thus, complete loss of Cdon pax2, are hyperactivated and the ventral quadrant of the retina function would overall reduce Hh dispersion leading to a HPE fails to develop55. These defects coincide with those of the cdon phenotype, as observed in humans carrying CDON mutant morphants, tempting us to anticipate that in the cavefish cdon alleles15 or in Cdon null mice11,20,25,57. Boc, Gas1 or LRP2 might expression might be altered. reinforce this Cdon function in vertebrates, as suggested by the Despite the clear rescue of cdon morphant phenotype by stronger phenotype observed in compound Cdon À / À ;Boc À / À inhibition of Hh signalling, we could not detect an evident mouse embryos24. Although we cannot completely exclude that expansion or upregulation of the expression of Hh target genes this compensation might explain the absence of midline defects in (ptc1 gli2, gli2b, gli3, boc and nkx2.1)incdon morphants, with the our Cdon morphants, we believe that our downregulation of Cdon exception of nkx2.2, which is broadly expressed in the anterior function in both zebrafish and chick bypasses this early function, hypothalamic region. Invariance of target gene expression after which is likely responsible for the reduced expression of Pax2 gain of Hh function has been reported in other studies, in which observed in the optic vesicle of Cdon À / À mouse embryos26. the phenotype could be nonetheless rescued by modifying the The incomplete downregulation of cdon in zebrafish has, activity of other Hh pathway components49,53. These however, the advantage of unveiling a later Cdon effect:

8 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE

Apical Basal Apical

Apical Cherry EGFP-GPI

Basal Basal Basal Apical α-Cherry α-Shh Merge FP Apical

Basal Apical Cdon Boc-EGFP Basal Apical Basal α -Cdon α-Shh Merge FP Basal Apical

Apical 8 ***

m) *** μ 6 ** FNIII(1-2) Δ Cdon Basal 4 α-Cdon α-Shh Merge

2

Apical Width of the basal end-foot ( 17 14 16 13 0

Cdon-FNIII(3) Cdon Cherry FNIII(3) FNIII(1-2)Δ α Δ α-Cdon Basal -Shh Merge Cdon Cdon

Figure 7 | Cdon and Boc promote morphological changes of the neuroepithelial basal side where Shh preferentially accumulates. (a–r) Confocal analysis of coronal sections at the level of HH14 chicken optic stalk (a–l) and HH10 neural tube (m–r) electroporated at HH8 with a construct carrying mCherry (a–c)Cdon(d–f), CdonDFnIII(1-2) (g–i), CdonDFnIII(3) (j–l), EGFP-GPI (m–o) or Boc-EGFP (p–r). Sections in (a–l) were immunostained with antibodies against Cherry (a)Cdon(d,g,j) and Shh (b,e,h,k). Images in (n,o) and (q,r) are high magnification views of cells shown in (m,p) respectively. Note how cells electroporated with Boc-EGFP (q,r, arrows), Cdon (d) or CdonDFnIII(1-2) (g) present an enlarged basal end-foot when compared with EGFP, mCherry or CdonDFnIII(3) neuroepithelial cells (n,o and a,j arrows). This enlarged end-foot is a preferential site of Shh accumulation (e,f,h,i). Note also the absence of Shh signal in the cells immediately surrounding the Cdon-positive cells (e,h, dotted line arrows). This distribution is not observed n cells expressing mCherry (b,c) or CdonDFnIII(3) (k,l). (s) Quantification of the basal end-foot width of neuroepithelial cells electroporated with Cdon and its deleted versions. The number of cases analysed for each data set is indicated in the respective column. The numbers of quantified cells are indicated in the graph labels and are as follows: Cherry, n ¼ 17; Cdon, n ¼ 14; CdonDFnIII(1-2), n ¼ 16 and CdonDFnIII(3), n ¼ 13. Error bars represent s.e.m. (**Po0.01, ***Po0.001; Student’s t-test). There is no statistical difference in the basal end-foot width between Cdon and CdonDFnIII(1-2) expressing cells (P ¼ 0.066, Student’s t-test) or between Cherry and CdonDFnIII(3) expressing cells (P ¼ 0.998; Student’s t-test). Scale bar, 10 mm. protecting the retinal neuroepithelium from Hh signalling. In binding—mimicked cdonATG morphant phenotype, whereas absence of Ptc, Cdon-mediated tight Hh binding would not lead removal of cdon Ptc-binding domain had no obvious effect on to signalling activation but rather to precluding Hh dispersion. eye development. Analysis of the effects of Cdon and its FNIII(2) Therefore, the same Ihog/Cdon function of tightly binding the and FNIII(3) deleted derivatives in localized rescue experiments available Hh may limit long-range or enhance signalling and on the endogenous Hh distribution led to similar conclu- depending on their expression, their distance to the Hh source sions. We show that neuroepithelial cells expressing Cdon or the concomitant presence of the Ptc receptor. accumulate Hh protein on their surface independently from the This Ptc1-independent decoy function of Cdon is supported by presence of the Ptc interacting domain, indicating that this Cdon the complementary expression pattern of Cdon and Ptc in the eye variant still effectively binds Hh, well in agreement with studies and other neural regions. Furthermore, the use of exon skipping with human CDON mutated alleles15 and with Ihog mutant MO demonstrated that CdonDFnIII(3)—deficient in Hh forms in Drosophila19. Cdon ectopic expression did not change

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 9 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 the fate of targeted cells into Hh-producing cells, as determined Embryo injections. The MOs used in this study are listed in Supplementary by ISH analysis (not shown). Thus, we propose that Cdon as well Table 1. The MOct and cMOct are standard control MOs from Gene Tools. MOs as Boc can tightly bind Hh, depleting the amount of morphogen were injected at concentrations listed in Supplementary Table 1 (1nl) by using a microinjector (Femtojet/ Eppendorf or IM-300/Narishige) in zebrafish embryos at present in the surrounding cells, as observed in our studies one-cell stage. To determine cdonATG MO efficiency, the cdonATG MO or MOct (Fig. 7). This strong accumulation prevents further diffusion of were co-injected into medaka fish embryos at one-cell stage with the RNA coding the ligand, thereby blocking pathway activation. GFP alone or with the cRNA (50 ng ml À 1) of a reporter construct (50cdon-GFP) 0 What is the destiny of the accumulated Hh protein is just a carrying the cdon 5 sequence upstream of the GFP sequence and the knockdown of GFP expression was analysed. The exon removal efficiency of cdonspl8, cdons- matter of speculation at the moment. In a simple view the protein pl11a, cdonspl11d and cdonspl14 MO was verified by RT–PCR using the cDNA of the could be endocytosed and degraded with a mechanism similar to respective morphant embryos at 24 h.p.f. The primers used to test effective exon that described for Glypican3 (ref. 58), Ptc34,59,60 or the Hh skipping are the following: cdonspl8MO, fwd: 50-TATCCATCCTGCGAGGT 0 0 0 spl11a spl11d interacting protein61. An alternative possibility comes from the TTG-3 , rev 5 -GACCTCATACAGGCTGGACG-3 ; cdon and cdon MOs fwd: 50-CCAATAAGAATCCCTCCA AAG-30, rev 50-CCAGCGTAACAGAATCT observation that Cdon and Boc expression is associated with the GAG-30; cdonspl14 MO fwd: 50-ACACTCCATCCAGCAATAAC-30, rev 50-TCCT presence of filopodial-like extensions at the basal side of CCATAAGCACATGACG-30. PCR products were run in agarose gels (1–2%) neuroepithelial cells. Although shorter, these extensions may be and visualized with FastRed (Biotium). The different bands were excised from the related to the cytonemes initially described in Drosophila as gel, ligated by T-cloning in the pSC-A vector (Stratagene) and sequenced. 62 Sequences are reported in Supplementary Fig. 8. cdonspl8 MO injection generated structures implicated in morphogen transport and recently three aberrant cdon mRNAs: band 2 carrying a deletion of 73 bp that causes a proven to be at the base of Hh long-range distribution in frame shift; band 3 carrying a 162 bp deletion and band 4 corresponding to the Drosophila6 and in the vertebrate limb bud8. Cytonemes are frame-shift deletion of exon 8 (Fig. 4c). cdonspl14 MO injection generated one aberrant cdon mRNAs: band 6 corresponding to the exon14 skipping (Fig. 4f). present in both Hh producing and receiving cells and are sp11a spl11d enriched in Ihog6 or Cdon8. It is possible that in cells that should cdon and cdon MOs co-injection generated two aberrant cdon mRNAs: band 8 carrying a deletion of 29 bp that causes a frame shift and band 9 not receive Hh signalling, as the presumptive retinal epithelium, corresponding to the exon 11 skipping (Fig. 4i). The p53 MO was used according Cdon-enriched extensions could serve to reshuffle Hh to the to Robu et al.43 nearby Hh responding optic stalk cells. In this way, the activity of overly dispersed Hh will be maximized by being reconvened at Tissue processing and immunochemistry. Fish, chicken and mouse embryos the responding cells farthest away from the Hh source, thus were fixed by immersion in 4% paraformaldehyde-phosphate buffer (wt/vol) reinforcing signalling activation in these cells. Further studies overnight at 4 °C. Embryos were then washed in phosphate buffer saline (PBS), are needed to establish if these extensions are specific features incubated in a 15% sucrose-PBS solution (wt/vol), embedded and frozen in a 7.5% gelatin in 15% sucrose solution (wt/vol). Cryostat sections were processed for IHC of Cdon/Boc expressing cells or are normally present in or for in situ hybridization (ISH). To detect endogenous Shh distribution, chick all neuroepithelial cells in labile forms undetectable with embryos were instead fixed in ethanol 95%/acetic acid 1% for 20 min. Cryostat conventional fluorescent techniques, which become stabilized by sections or whole embryos were stained by a standard protocol using antibodies the expression of these proteins, as suggested in Drosophila6. against the following antigens: Cdon (1:100, R&D), Pax6 (1:500, Covance), Pax2 Notably, basolateral protrusions have been very recently (1:500, Zymed), Fluorescein (1:500, Roche), Shh (1:500, 5E1 Hybridoma bank), HA 63 (1:250, Sigma). For Hh and Pax2 immunostaining in zebrafish, we adapted the described in the mouse . protocol for tissue slices69 by boiling the embryos directly in citrate buffer 10 mM In Drosophila, Boi and Ihog have a redundant function in Hh at 110 °C during 5 min. Embryos were thereafter washed in PBS and incubated for pathway activation19 and can sequester and titrate the amount of two days in the following primary antibodies: anti-Shh (1:50, R&D) and anti-Pax2 ligand available for Ptc binding, thus limiting long-range (1:50, Zymed). Incubation with appropriate secondary antibodies was performed with standard procedures. signalling19,37,38,64. Ihog is also involved in cytoneme-based Hh signalling6. Thus, at least Ihog seems to have both a positive and negative effect on Hh signalling activation38. Our study, together In situ hybridization. Chicken embryos were hybridized using digoxigenin- or fluorescein-UTP-labelled antisense riboprobes for pax2 and cdon. Probes were with the observations that Cdon, Boc and Gas1 have redundant visualized with NBT/BCIP (dark blue). Zebrafish embryos were hybridized in toto 17,20,22,24,57 functions as Hh co-receptors , indicates that this dual by standard procedures using the following digoxigenin-UTP-labelled antisense activity is conserved in the vertebrate Cdon homologue and riboprobes: tbx5.1, fgf8, pax2.1, ptc1, shh, nkx2.1, nkx2.2 and rdlh10a. Chicken perhaps also in Boc. Indeed, Boc overexpression correlates with embryos were hybridized in toto by standard procedures using the following digoxigenin-UTP-labelled antisense riboprobes: pax2 and cdon. The zebrafish boc the presence of filopodial-like extensions at the basolateral side of and cdon coding sequences were obtained by RT–PCR from cDNA of embryos the neuroepithelium (Fig. 7) and its expression seems to protect collected at different developmental stages using the following primers (boc: fwd: mouse ipsilateral projecting retinal ganglion cells from Shh 50-CATCGATCCTTTCAATGCAAG-30, rev: 50-GGGAGTATTCTTGTTTCAT 65 CCA-30; cdon: fwd: 50-CATCGGGAGAATGTGTTTCG-30, rev: 50-TCCACCAAT activity . Similarly, high levels of Boc expressed in the zebrafish 0 jaw have been proposed to bind Shh and limit the ability of Hh to ATCTTCATTCG-3 ) and cloned in pDrive vector (Qiagen). activate chondrogenesis and proliferation in this tissue28. Whether Cdon and Boc participate in Hh release/transport, as mRNA and cDNA synthesis and cloning. Embryos were injected with different Ihog does, remains to be determined. Furthermore, it might be splicing MOs at one-cell stage and at 24 h.p.f. were frozen in dry ice. RNA pur- ification was performed with PureLink miniKit (Invitrogen) and cDNA was syn- worth exploring whether Cdon has Hh unrelated function in the thesized with iScript cDNA synthesis Kit (Bio-Rad). 50cdon-GFP construct was developing eye. Indeed independent of its Hh-related activity, obtained by fusing the RT–PCR amplified cdon coding sequence upstream of the Cdon has been shown to interact with N-cadherin66 which, in GFP sequence in a pCS2 vector (pCS2-50cdon-GFP). Capped RNA (cRNA) was 0 turn, is relevant to proper optic fissure closure67. synthesized using pCS2-5 cdon-GFP linearized with NotI as a template and the mMessage mMachine SP6 Kit (Ambion). The resulting mRNA was purified using RNeasy Mini Kit (Qiagen). 50cdon-GFP mRNA was injected (1nl) at 80 ng ul À 1.To generate the UAS constructs, mCdon-HA was amplified from pCIG-mCdon using Methods the Expand High Fidelity PCR system (Roche) and the following primers fwd: Maintenance of fish lines and chicken embryos. Adult zebrafish (Danio rerio) 50-ATAGAATTCACGTGCTGGTTATTGTGCTG-30, rev: 50-ATAGAATTCAGC and the Tg(rx3:Gal4-VP16) line48 were maintained at 28.5 °C on a 14/10 h light/ GGCTTCGGCCAGTAACG-30. The PCR product was cloned with T-cloning in dark cycle. Embryos (AB/Tu or WIK strains) were raised at 28 °C and staged pSC-A (Stratagene). The vector was digested with EcoRI/FbaI and the band according to hours post fertilization (h.p.f.) and morphology68. Embryos were corresponding to mCdon-HA was cloned into the pCS2 vector digested with growth in E3 medium (NaCl, 5 mM; KCl, 0.17 mM; CaCl2, 0.33 mM; MgSO4, EcoRI. The pCS2-mCdon vector was verified by restriction analysis with EcoRV/ 0.33 mM, 5.10% Methylene Blue). We used chicken embryos (Gallus gallus XhoI digestion and by DNA sequencing. Versions of Cdon lacking FnIII-2 or the domestica) of the White Leghorn breed raised in the Santa Isabel farm FnIII-3 were generated by PCR amplification with a PfuUltra DNA polymerase (Co´rdoba, Spain). The procedures used in the study were approved by the Ethical (Agilent) from the pCS2-mCdon vector using the following 50phosphorilated Committee for Animal Experimentation of the Consejo Superior de Investigaciones primers 50-CAGGTGGCCGGCTTCCCAAATC-30,50-ATGCCTGGAGGAATC Cientificas (CSIC). CGTAAG-30 for deletion of FNIII(2) and 50-GGAGCTTCCGACTATCCCGTG-30,

10 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE

50-CACCTGGTAAGGACGGGATGC-30 for deletion of FNIII(3). The amplified Digital images were obtained using DFC500, DFC350 FX cameras (Leica) or products were ligated (Ligase, Roche) overnight at 16 °C, digested with DpnI confocal microscopy (Zeiss) and processed with Photoshop CS5 or ImageJ (Fiji) enzyme to remove the non-amplified DNA and then transformed. The pCS2- software. mCdon, containing an HA tag, and its derivatives (pCS2-CdonDFnIII(2) and pCS2-CdonDFnIII(3)) were transfected in HEK cells and the expected products were verified by immunostaining and western blot analysis with anti-Cdon and Quantification and statistical analysis. For pax2.1 quantifications, embryos were anti-HA antibodies. pCS2-mCdon, pCS2-CdonDFnIII(2) and pCS2-CdonDF- mounted in glycerol 85% and photographed in lateral views at  160 to quantify nIII(3) vectors were digested with ClaI and the product was treated with Klenow the area of the pax2.1-positive optic stalk domain. Comparison of nkx2.2 and polymerase for blunting DNA ends. The product was digested with NotI/SalI and nkx2.1-positive domains in control and morphant embryos was performed using run in an agarose gel. The band of approximately 4 Kb was excised from the gel, dorsal view images of whole embryos and determining the mediolateral extent of purified and ligated in pBR-Tol2-UAS:MCS-UAS:GFP previously digested with the signal. Quantitative analysis of Shh levels in electroporated cells was performed EcoRV/NotI (in short, UAS::GFP, which was kindly provided by Dr Masa Tada). using confocal images of transversal sections of control and experimental embryos The obtained vectors UAS::Cdon, UAS::CdonDFnIII(2) and UAS::CdonDFnIII(3) at the level of the optic stalk. In each one of the selected sections (one section per were verified by restriction analysis with KpnI and EcoRV/BglII digestion, PCR embryo, 7–14 embryos per case were analysed), Shh levels were measured in two amplification and sequencing. Efficient expression and membrane localization of adjacent regions (proximal and distal) outside of the endogenous Shh source. These the constructs were verified by DNA microinjection in the rx3::Gal4 transgenic line values were plotted separately. Quantitative analysis of width of the basal end-foot and a-HA immunostaining at 20 h.p.f. (Supplementary Fig. 10). area was performed using confocal images of Cherry, Cdon or deleted derivatives overexpressing cells. Quantifications were performed using ImageJ (NIH). Data analyses were performed with the IBM SPSS statistic software. All the data sets Deletion constructs of Cdon and Boc. pCIG-mCdon-HA (pCIG-mCdon), presented a normal distribution (Kolmogorov–Smirnov test). Analysis of the data pCIG-rCdon-FnIIID12-HA (CdonDFnIII(1-2)) and pCIG-mCdon-FnIIID3-HA was performed using t-test for two groups and analysis of variance for larger (CdonDFnIII(3)) vectors were kindly provided by Dr B. Allen17. pEGFP-mBoc groups. Statistical differences between pools of treated (MO, drugs, overexpression) expression vector (Boc-EGFP) was kindly provided by Dr Ami Okada65. Versions and control embryos were determined with the Pearson’s w2 test. In all graphs the of Boc lacking FnIII-2 or the FnIII-3 were generated by PCR amplification with a error bars indicate s.e.m. PfuUltra DNA polymerase (Agilent) from the pEGFP-mBoc plasmid using the following primers: Boc-GFPDFnIII-2 fwd: 50-GTGTCAGGCTACAGTGGC-30, rev: 50-ATGGTCAGGCTGGCTGCT-30; Boc-GFPDFnIII-3 fwd: 50-TTTTCTGGTCA References GCCTGGA-30, rev: 50-CCTCTCATATACACGGCC-30. The amplified products 1. Ingham, P. W., Nakano, Y. & Seger, C. Mechanisms and functions of Hedgehog were ligated (Ligase, Roche) overnight at 16 °C, digested with DpnI enzyme to signalling across the metazoa. Nat. Rev. Genet. 12, 393–406 (2011). remove the non-amplified DNA and then transformed. 2. Ryan, K. E. & Chiang, C. Hedgehog secretion and signal transduction in vertebrates. J. Biol. Chem. 287, 17905–17913 (2012). 3. Bornemann, D. J., Duncan, J. E., Staatz, W., Selleck, S. & Warrior, R. Abrogation Chicken electroporation. Fertilized chicken eggs were incubated at 38 °C to reach of heparan sulfate synthesis in Drosophila disrupts the Wingless, Hedgehog and the desired stage. A carboxyfluorescein-tagged70 chicken-specific MO (cCdonATGMO) was designed to be complementary to the translation start site Decapentaplegic signaling pathways. Development 131, 1927–1938 (2004). sequence of Cdon (NCBI Reference Sequence: XM_417853.4). The MO efficiently 4. Callejo, A. et al. Dispatched mediates Hedgehog basolateral release to form the reduced the levels of the endogenous protein (Supplementary Fig. 11). The selected long-range morphogenetic gradient in the Drosophila wing disk epithelium. vectors (1.5 mg ml À 1), cMOct or cCdonATG were injected in ovo in the eye Proc. Natl Acad. Sci. USA 108, 12591–12598 (2011). primordium of HH8 chick embryos right at the appearance of cdon expression in 5. Ma, Y. et al. Hedgehog-mediated patterning of the mammalian embryo the forming vesicle. The electrodes were placed at the head sides. Focal electrical requires transporter-like function of dispatched. Cell 111, 63–75 (2002). discharges were applied with the following conditions: voltage, 14V; pulses, 5; 6. Bischoff, M. et al. Cytonemes are required for the establishment of a normal interval, 300 ms and pulse length, 50 ms. Thereafter, eggs were sealed and Hedgehog morphogen gradient in Drosophila epithelia. Nat. Cell. Biol. 15, incubated at 38 °C until embryos reached HH14, when cdon mRNA localizes to the 1269–1281 (2013). entire retina (Supplementary Fig. 11). Embryos were then selected according to 7. Rojas-Rios, P., Guerrero, I. & Gonzalez-Reyes, A. Cytoneme-mediated delivery GFP expression and fixed for further processing. of hedgehog regulates the expression of bone morphogenetic proteins to maintain germline stem cells in Drosophila. PLoS Biol. 10, e1001298 (2012). 8. Sanders, T. A., Llagostera, E. & Barna, M. Specialized filopodia direct long- Rescue experiments ATG . Cdon MO and UAS::GFP, UAS::Cdon, UAS::CdonDF- range transport of SHH during vertebrate tissue patterning. Nature 497, D nIII(2), and UAS::Cdon FnIII(3) DNAs were co-injected in the transgenic 628–632 (2013). line Tg(rx3:Gal4-VP16). DNA (35 pg) was injected into the cell of one-cell stage 9. Denef, N., Neubuser, D., Perez, L. & Cohen, S. M. Hedgehog induces opposite embryos, whereas CdonATGMO (1nl of at 160 uM) was introduced in a second changes in turnover and subcellular localization of patched and smoothened. round of injections into the yolk at the two–four-cells stage. Embryos were Cell 102, 521–531 (2000). let to develop for 26–28 h in E3 medium at 28 °C. Embryos showing a homo- 10. Taipale, J., Cooper, M. K., Maiti, T. & Beachy, P. A. Patched acts catalytically to geneous expression of GFP in the retina were selected and further processed for pax2.1 ISH. suppress the activity of Smoothened. Nature 418, 892–897 (2002). 11. Cole, F. & Krauss, R. S. Microform holoprosencephaly in mice that lack the Ig superfamily member Cdon. Curr. Biol. 13, 411–415 (2003). Cell transfection. Sub-confluent Human Embryonic Kidney 293T (HEK293) cells 12. Kang, J. S. et al. CDO: an oncogene-, serum-, and anchorage-regulated member were transiently co-transfected with the pCIG-Cdon-HA and derivatives or with of the Ig/fibronectin type III repeat family. J. Cell Biol. 138, 203–213 (1997). the pEGFP-mBoc or derivative-plasmids using the FuGENE HD Transfection 13. Kang, J. S., Mulieri, P. J., Hu, Y., Taliana, L. & Krauss, R. S. BOC, an Ig Reagent (Roche). After 48 h, cells were visualized by fluorescence in a confocal superfamily member, associates with CDO to positively regulate myogenic microscope or scraped in lysis buffer for western blot analysis. differentiation. EMBO J. 21, 114–124 (2002). 14. Sanchez-Arrones, L., Cardozo, M., Nieto-Lopez, F. & Bovolenta, P. Cdon and Western blotting. HEK293 cell or tissue samples (electroporated chicken ) Boc: Two transmembrane proteins implicated in cell-cell communication. Int. J. were collected or dissected, treated with lysis buffer (150 mM NaCl, 1% NP40, Biochem. Cell Biol. 44, 698–702 (2012). 50 mM Tris pH8 ) and denatured in protein loading buffer (50 mM Tris–HCl pH 15. Bae, G. U. et al. Mutations in CDON, Encoding a Hedgehog receptor, result in 6.8, 2% SDS, 10% glycerol, 1% b-mercaptoethanol, 12.5 mM EDTA and 0.02% Holoprosencephaly and defective interactions with other hedgehog receptors. bromophenol blue). Proteins were resolved by 6% SDSP-gels, blotted onto a PVDF Am. J. Hum. Genet. 89, 231–240 (2011). membrane, blocked in 5% milk in PBS Tween-20 for 1 h, incubated with Cdon 16. McLellan, J. S. et al. The mode of Hedgehog binding to Ihog homologues is not polyclonal (R&D system) or HA polyclonal antibody (Sigma) in blocking conserved across different phyla. Nature 455, 979–983 (2008). buffer overnight at 4 °C, washed four times with PBST, and incubated for 1 h with 17. Tenzen, T. et al. The cell surface membrane proteins Cdo and Boc are peroxidase-conjugated secondary antibody. Labelled proteins were detected with components and targets of the Hedgehog signaling pathway and feedback the chemiluminescence reagent ECL (Amersham Biosciences). network in mice. Dev. Cell. 10, 647–656 (2006). 18. Yao, S., Lum, L. & Beachy, P. The ihog cell-surface proteins bind Hedgehog and mediate pathway activation. Cell 125, 343–357 (2006). Treatment with Hh signalling inhibitor . Cyclopamine (Calbiochem) was used to 19. Zheng, X., Mann, R. K., Sever, N. & Beachy, P. A. Genetic and biochemical antagonize Hh signalling. Embryos at 90% epiboly were incubated in E3 medium definition of the Hedgehog receptor. Genes Dev. 24, 57–71 (2010). containing 100 mM Cya dissolved in dimethyl sulphoxide at 28 °C. Embryos were 20. Allen, B. L., Tenzen, T. & McMahon, A. P. The Hedgehog-binding proteins fixed in 4% paraformaldehyde at 28 h.p.f. Gas1 and Cdo cooperate to positively regulate Shh signaling during mouse development. Genes Dev. 21, 1244–1257 (2007). Imaging. Sections were analysed with a DM or confocal microscope, whereas 21. Martinelli, D. C. & Fan, C. M. Gas1 extends the range of Hedgehog action by whole embryos were observed with a stereomicroscope (Leica Microsystems). facilitating its signaling. Genes Dev. 21, 1231–1243 (2007).

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 11 & 2014 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272

22. Allen, B. L. et al. Overlapping roles and collective requirement for the 52. Lee, J. et al. An ENU mutagenesis screen in zebrafish for visual system mutants coreceptors GAS1, CDO, and BOC in SHH pathway function. Dev. Cell 20, identifies a novel splice-acceptor site mutation in patched2 that results in 775–787 (2011). Colobomas. Invest. Ophthalmol. Vis. Sci. 53, 8214–8221 (2012). 23. Izzi, L. et al. Boc and Gas1 each form distinct Shh receptor complexes 53. Heisenberg, C. P., Brennan, C. & Wilson, S. W. Zebrafish aussicht mutant with Ptch1 and are required for Shh-mediated cell proliferation. Dev. Cell 20, embryos exhibit widespread overexpression of ace (fgf8) and coincident defects 788–801 (2011). in CNS development. Development 126, 2129–2140 (1999). 24. Zhang, W., Hong, M., Bae, G. U., Kang, J. S. & Krauss, R. S. Boc modifies the 54. Menuet, A., Alunni, A., Joly, J. S., Jeffery, W. R. & Retaux, S. Expanded holoprosencephaly spectrum of Cdo mutant mice. Dis. Model Mech. 4, 368–380 expression of Sonic Hedgehog in Astyanax cavefish: multiple consequences (2011). on forebrain development and evolution. Development 134, 845–855 25. Hong, M. & Krauss, R. S. Cdon mutation and fetal ethanol exposure synergize (2007). to produce midline signaling defects and holoprosencephaly spectrum disorders 55. Yamamoto, Y., Stock, D. W. & Jeffery, W. R. Hedgehog signalling controls eye in mice. PLoS Genet. 8, e1002999 (2012). degeneration in blind cavefish. Nature 431, 844–847 (2004). 26. Zhang, W. et al. Ocular abnormalities in mice lacking the immunoglobulin 56. Ingham, P. W. Hedgehog signaling. Cold Spring Harb. Perspect. Biol. 4, a011221 superfamily member Cdo. FEBS J. 276, 5998–6010 (2009). (2012). 27. Zhang, W. et al. Cortical thinning and hydrocephalus in mice lacking the 57. Zhang, W., Kang, J. S., Cole, F., Yi, M. J. & Krauss, R. S. Cdo functions at immunoglobulin superfamily member CDO. Mol. Cell. Biol. 26, 3764–3772 multiple points in the Sonic Hedgehog pathway, and Cdo-deficient mice (2006). accurately model human holoprosencephaly. Dev. Cell 10, 657–665 (2006). 28. Bergeron, S. A., Tyurina, O. V., Miller, E., Bagas, A. & Karlstrom, R. O. Brother 58. Capurro, M. I. et al. Glypican-3 inhibits Hedgehog signaling during of cdo (umleitung) is cell-autonomously required for Hedgehog-mediated development by competing with patched for Hedgehog binding. Dev. Cell 14, ventral CNS patterning in the zebrafish. Development 138, 75–85 (2011). 700–711 (2008). 29. Schwarz, M. et al. Spatial specification of mammalian eye territories by 59. Chen, Y. & Struhl, G. Dual roles for patched in sequestering and transducing reciprocal transcriptional repression of Pax2 and Pax6. Development 127, Hedgehog. Cell 87, 553–563 (1996). 4325–4334 (2000). 60. Torroja, C., Gorfinkiel, N. & Guerrero, I. Patched controls the Hedgehog 30. Ekker, S. C. et al. Patterning activities of vertebrate hedgehog proteins in the gradient by endocytosis in a dynamin-dependent manner, but this developing eye and brain. Curr. Biol. 5, 944–955 (1995). internalization does not play a major role in signal transduction. Development 31. Macdonald, R. et al. The Pax protein Noi is required for commissural axon 131, 2395–2408 (2004). pathway formation in the rostral forebrain. Development 124, 2397–2408 61. Chuang, P. T. & McMahon, A. P. Vertebrate Hedgehog signalling modulated by (1997). induction of a Hedgehog-binding protein. Nature 397, 617–621 (1999). 32. Macdonald, R. et al. Midline signalling is required for Pax gene regulation and 62. Ramirez-Weber, F. A. & Kornberg, T. B. Cytonemes: cellular processes that patterning of the eyes. Development 121, 3267–3278 (1995). project to the principal signaling center in Drosophila imaginal discs. Cell 97, 33. Chiang, C. et al. Cyclopia and defective axial patterning in mice lacking Sonic 599–607 (1999). hedgehog gene function. Nature 383, 407–413 (1996). 63. Williams, M., Yen, W., Lu, X. & Sutherland, A. Distinct apical and basolateral 34. Lee, J. et al. Zebrafish blowout provides genetic evidence for Patched1-mediated mechanisms drive planar cell polarity-dependent convergent extension of the negative regulation of Hedgehog signaling within the proximal optic vesicle of mouse neural plate. Dev. Cell 29, 34–46 (2014). the vertebrate eye. Dev. Biol. 319, 10–22 (2008). 64. Bilioni, A. et al. Balancing hedgehog, a retention and release equilibrium given 35. Mulieri, P. J., Okada, A., Sassoon, D. A., McConnell, S. K. & Krauss, R. S. by Dally, Ihog, Boi and shifted/dWif. Dev. Biol. 376, 198–212 (2012). Developmental expression pattern of the cdo gene. Dev. Dyn. 219, 40–49 65. Sanchez-Arrones, L. et al. Shh/Boc signaling is required for sustained (2000). generation of ipsilateral projecting ganglion cells in the mouse retina. 36. Bilioni, A. et al. Balancing Hedgehog, a retention and release equilibrium given J. Neurosci. 33, 8596–8607 (2013). by Dally, Ihog, Boi and shifted/DmWif. Dev. Biol. 376, 198–212 (2013). 66. Lu, M. & Krauss, R. S. N-cadherin ligation, but not Sonic hedgehog binding, 37. Hartman, T. R. et al. Drosophila Boi limits Hedgehog levels to suppress follicle initiates Cdo-dependent p38alpha/beta MAPK signaling in skeletal myoblasts. stem cell proliferation. J. Cell Biol. 191, 943–952 (2010). Proc. Natl Acad. Sci. USA 107, 4212–4217 (2010). 38. Yan, D. et al. The cell-surface proteins Dally-like and Ihog differentially 67. Masai, I. et al. N-cadherin mediates retinal lamination, maintenance of regulate Hedgehog signaling strength and range during development. forebrain compartments and patterning of retinal neurites. Development 130, Development 137, 2033–2044 (2010). 2479–2494 (2003). 39. Torres, M., Go´mez-Pardo, E. & Gruss, P. Pax2 contributes to inner ear 68. Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B. & Schilling, T. F. patterning and optic nerve trajectory. Development 122, 3381–3391 (1996). Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310 40. Concordet, J. P. et al. Spatial regulation of a zebrafish patched homologue (1995). reflects the roles of sonic hedgehog and protein kinase A in neural tube and 69. Liu, H. X. et al. Multiple Shh signaling centers participate in fungiform papilla patterning. Development 122, 2835–2846 (1996). and taste bud formation and maintenance. Dev. Biol. 382, 82–97 (2013). 41. Chang, L., Blain, D., Bertuzzi, S. & Brooks, B. P. Uveal coloboma: clinical and 70. Voiculescu, O., Papanayotou, C. & Stern, C. D. Spatially and temporally basic science update. Curr. Opin. Ophthalmol. 17, 447–470 (2006). controlled electroporation of early chick embryos. Nat. Protoc. 3, 419–426 42. Sehgal, R., Karcavich, R., Carlson, S. & Belecky-Adams, T. L. Ectopic Pax2 (2008). expression in chick ventral optic cup phenocopies loss of Pax2 expression. Dev. Biol. 319, 23–33 (2008). 43. Robu, M. E. et al. p53 activation by knockdown technologies. PLoS Genet. 3, Acknowledgements e78 (2007). We are in debt with P. Esteve and F. Cavodeassi for support and critically reading the 44. Lupo, G. et al. Dorsoventral patterning of the Xenopus eye: a collaboration of manuscript. We thank the CBMSO Image analysis platform and the Zebrafish Platform Retinoid, Hedgehog and FGF receptor signaling. Development 132, 1737–1748 at the CABD, Seville for their invaluable help. We are also grateful to Masa Tada, Ami (2005). Okada, Ben Allen and Adi Inbal for reagents and material and to Noemi Tabanera, 45. Dente, L. et al. Cloning and developmental expression of zebrafish pdzrn3. Int. Alfonso Gutie´rrez and Isidro Dompablo for the excellent technical support and fish J. Dev. Biol. 55, 989–993 (2011). caring. This work was supported by grants from the Spanish Ministerio de Economı´ay 46. Lupo, G. et al. Retinoic acid receptor signaling regulates choroid fissure closure Competitividad (MINECO) (BFU2010-16031), Comunidad Auto´noma de Madrid through independent mechanisms in the ventral optic cup and periocular (CAM, CELL-DD S2010/BMD-2315), the Centro de Investigacio´n Biome´dica en Red de mesenchyme. Proc. Natl Acad. Sci. USA 108, 8698–8703 (2011). Enfermedades Raras (CIBERER) del Instituto Carlos III (ISCIII) to P.B., MINECO 47. Isaac, A. et al. Tbx genes and limb identity in chick embryo development. (BFU2011-25987) to I.G. and an Institutional Grant from the Fundacion Ramon Areces. Development 125, 1867–1875 (1998). M.J.C. was supported by a predoctoral contract from the MINECO (BES-2008-005457). 48. Weiss, O., Kaufman, R., Michaeli, N. & Inbal, A. Abnormal vasculature L.S.-A. holds a postdoctoral contract from the CSIC (JAEDOC-012) and Fondo Social interferes with optic fissure closure in lmo2 mutant zebrafish embryos. Dev. Europeo. Work in the Wilson group was supported by grants from the MRC G0501487 Biol. 369, 191–198 (2012). and G0900994 (G.G. and S.W.W.) and Wellcome Trust (089227/Z09/Z S.W.W.). 49. Sanek, N. A., Taylor, A. A., Nyholm, M. K. & Grinblat, Y. Zebrafish zic2a patterns the forebrain through modulation of Hedgehog-activated gene expression. Development 136, 3791–3800 (2009). Author contributions 50. Holtz, A. M. et al. Essential role for ligand-dependent feedback antagonism of M.J.C. participated in all experiments presented in this work, analysed data and vertebrate hedgehog signaling by PTCH1, PTCH2 and HHIP1 during neural contributed to experimental design and manuscript writing. L.S.-A. performed studies in patterning. Development 140, 3423–3434 (2013). chick embryos and analysed the data. A.S. performed histological and confocal work 51. Zhang, X. M. & Yang, X. J. Temporal and spatial effects of Sonic hedgehog and analysed the data. C.S.-C. performed experiments related to Boc function. signaling in chick eye morphogenesis. Dev. Biol. 233, 271–290 (2001). S.W.W., G.G. and I.G. contributed to experimental design and revised the manuscript.

12 NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications & 2014 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5272 ARTICLE

Artwork was generated by M.J.C. and L.S.A. P.B. directed the work, analysed data and How to cite this article: Cardozo, M-J. et al. Cdon acts as a Hedgehog decoy wrote the manuscript. receptor during proximal-distal patterning of the optic vesicle. Nat. Commun. 5:4272 doi: 10.1038/ncomms5272 (2014). Additional information Supplementary Information accompanies this paper at http://www.nature.com/ This work is licensed under a Creative Commons Attribution 4.0 naturecommunications International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise Competing financial interests: The authors declare no competing financial interests. in the credit line; if the material is not included under the Creative Commons license, Reprints and permission information is available online at http://npg.nature.com/ users will need to obtain permission from the license holder to reproduce the material. reprintsandpermissions/ To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 5:4272 | DOI: 10.1038/ncomms5272 | www.nature.com/naturecommunications 13 & 2014 Macmillan Publishers Limited. All rights reserved.